Filozofická fakulta Univerzity Palackého

Katedra anglistiky a amerikanistiky

Serial Verb Constructions

(Bakalářská práce)

2015 Timotej Šefčovič

Serial Verb Constructions (Bakalářská práce)

Autor: Timotej Šefčovič Studijní obor: Anglická filologie - Japonská filologie Vedoucí práce: Joseph Embley Emonds Prof. Počet stran: 75 Počet znaků: 88 063 (bez mezer) Olomouc 2015

I declare that I elaborated this paper myself and that I mentioned the absolute list of works cited.

In Olomouc on the day ………………………………………………...

Acknowledgement

However various or complex the language is, there will never be enough words for those who have been supporting me or even contributed directly to successful completing of this research.

My modest and humble “thank you” belongs to the teachers from the Japanese section of the Asian department at Palacký University for the knowledge of Japanese; to the members of the English department who have supported me in my enthusiasm for linguistics; to the teachers at Utsunomiya University, among other things, also for the knowledge of Korean; to my friends and especially to my family who have helped me in many other ways.

Above all, my deepest gratitude belongs to my supervisor Prof. Joseph Emonds especially for his permanent support and valuable advice that helped me to successfully finish this bachelor thesis. Abstract This thesis deals with morphological and syntactic properties of Serial Verb Constructions (SVCs). The SVCs can be found across languages from different parts of the world, especially in Bantu languages in Central Africa, and some Asian languages. However, there are also languages (e.g. Germanic ones) that seem to lack these constructions created by the process of serialization.

The objective of this thesis is to first analyse their structure and sum up their morphological and syntactic properties that were previously demonstrated by other linguists. Secondly, a new proposal concerning the fact that SVCs can be also found in Japanese is made. In order to fulfil this task, a closer look is held on Korean whose word order is identical to that in Japanese.

The final part of this thesis questions previous approaches to SVCs in Japanese, especially concerning their terminology. Afterwards, the outcome of this thesis suggests whether the previous approaches should be preserved or whether they should be updated and unified under a new terminology. Having examined the properties of SVCs in Korean and Japanese, we move back to the syntax of English, where English bare infinitives are compared to the already discussed SVCs.

Key words SVCs, verb, English bare infinitives, Japanese, Korean, compound, serialization, transitivity, argument Anotace Tato práce se zabývá morfologickými a syntaktickými vlastnostmi konstrukcí sériových sloves. Sériová slovesa lze nalézt v různých jazycích z různých částí světa, zejména v bantu jazycích ve střední Africe nebo v jiných asijských jazycích. Nicméně, existují i jazyky, například germánské, pro které nejsou tyto konstrukce vytvořené procesem serializace běžné.

Cílem této práce je nejprve analyzovat jejich strukturu a shrnout jejich morfologické a syntaktické vlastnosti. Všechny už byly dříve analyzovány jinými lingvisty. Za druhé je tu vyslovena teorie, že sériová slovesa se nacházejí i v japonštině. Pro splnění tohoto úkolu je využito korejštiny, která má slovosled stejný jako japonština.

Závěrečná část této práce analyzuje předcházející přístupy k sériovým slovesům v japonštině, zejména jejich terminologii. Poté tato práce navrhuje, zda by měla být zachována předchozí terminologie, nebo zda by měla být aktualizována a sjednocena pod jedním termínem. Poté se dostáváme zpátky k angličtině, kde budou porovnány už vzpomínaná sériová slovesa s anglickými holými infinitivy.

Klíčová slova sériová slovesa, sloveso, holý infinitiv, japonština, korejština, složenina, serializace, tranzitivita, argument

TABLE OF CONTENTS

1 INTRODUCTION ...... 9

2 SERIAL VERB CONSTRUCTIONS: THE ENGLISH SUBCASE ...... 11

2.1 WHAT ARE (NOT) SERIAL VERB CONSTRUCTIONS ...... 11

2.2 ENGLISH COMPOUND VERBS ...... 12

2.3 ENGLISH BARE INFINITIVES ...... 16

2.4 CONCLUSION: ENGLISH V-V AND SVCS ...... 18

3 GENERAL PROPERTIES OF SERIAL VERB CONSTRUCTIONS ...... 19

3.1 INTRODUCTION ...... 19

3.2 NO OVERT MARKER ...... 20

3.3 SUBJECT SHARING ...... 21

3.4 TENSE SHARING ...... 22

3.5 NEGATION ...... 22

3.6 HEADEDNESS ...... 23

4 JAPANESE SERIAL VERB CONSTRUCTIONS ...... 25

4.1 INTRODUCTION ...... 25

4.2 NO OVERT MARKER ...... 26

4.3 SUBJECT SHARING ...... 28

4.4 TENSE SHARING ...... 29

4.5 NEGATION ...... 30

4.6 HEADEDNESS ...... 31

4.7 CONCLUSION CONCERNING SVCS IN JAPANESE ...... 34

5 SYNTAX AND TRANSITIVITY OF SERIAL VERB CONSTRUCTIONS ...... 36

5.1 ARGUMENT STRUCTURE ...... 36

5.2 INTRANSITIVE V1 + INTRANSITIVE V2 ...... 37

5.3 INTRANSITIVE V1 + TRANSITIVE V2 ...... 38

5.4 TRANSITIVE V1 + TRANSITIVE V2 ...... 42

5.5 TRANSITIVE V1 + INTRANSITIVE V2 ...... 44

5.6 TRANSITIVE V1 + DITRANSITIVE V2 ...... 46

5.7 CONCLUDING REMARKS ...... 48

6 TERMINOLOGY GAPS ...... 49 6.1 A PREVIOUS APPROACH TO SERIAL VERB CONSTRUCTIONS ...... 49

6.2 A COMPOUND VERB ...... 49

6.3 TWO TYPES OF “COMPOUND VERBS” ...... 52

6.3.1 “SYNTACTIC COMPOUND VERBS” ...... 52

6.3.2 “LEXICAL COMPOUND VERBS” ...... 55

6.4 CONCLUSION CONCERNING KAGEYAMA’S TERMINOLOGY ...... 57

7 A CROSS-LINGUISTIC COMPARISON IN ENGLISH, JAPANESE, KOREAN ...... 59

7.1 GAPPING ...... 60

7.2 STRESS ASSIGNMENT ...... 63

7.3 CONCLUSION CONCERNING ENGLISH BARE INFINITIVES ...... 65

8 SUMMARY ...... 67

9 ČESKÉ RESUMÉ ...... 69

10 REFERENCES ...... 72

10.1 WORKS CITED ...... 72

10.2 WORKS CONSULTED ...... 73

Abbreviations and Symbols Used in This Thesis

Acc accusative marker Dat dative marker Dec declarative Ditran ditransitive e.g. for example etc. and other things Gen genitive marker Hon honorific marker i.e. that is/that means Intran intransitive L linker N noun NP noun phrase Neg negation marker Nom nominative marker Pass passive Past past tense marker Pres present tense marker SSH Subject Sharing Hypothesis SVC serial verb construction TE Japanese TE-form Top topic marker Tran transitive V verb V1 first verb V2 second verb VP verb phrase

1 INTRODUCTION

This thesis deals with the phenomenon of Serial Verb Constructions (SVCs). The SVC is a very interesting syntactic structure that consists of various V-V combinations. Although this term might be very common for speakers of some languages, there surely are those readers who have never been in touch with such a term. For this reason, the first part of this thesis is focused on theoretical background, where the basic concepts and the notion of SVCs are illustrated. Later, a bit more specific approaches are taken, provided with adequate examples from various languages,

In this thesis, the following objectives are proposed:

§ to support the idea that although the definitions of SVCs might vary among linguists, there are some properties of these constructions which are common to every language that is said to have these constructions.

§ to point out an idea that SVCs might be, to some extent, found also in English bare infinitives.

§ based on the contrastive analysis of Korean SVCs, to claim that Japanese is also one of the languages, which have SVCs. Consequently, to analyse these structures and list all possible combinations of verbs in V-V constructions, based on their transitivity.

§ to propose an idea that the traditional approach to Japanese V-V constructions is out-dated, which has resulted in terminology clashes/gaps. Therefore the terminology concerning compound/serial verbs should be revised.

This thesis is consisting of five thematic parts. Firstly a more general approach is taken to the SVC phenomenon. In order to understand how these structures look

9 and how they function, examples of different sequences of verbs are illustrated, such as English compound verbs (Lieber 1983).

Having examined the behaviour of verb series in English, a closer look is taken on SVCs found in Korean. At this point, a comparative study of morphological, syntactic, and semantic properties is undertaken (Baker 1989, Collins 1993, Lee 1992, among others). As the outcome of this comparative study, five universal properties of SVCs are demonstrated with a special look at the importance of subject/object sharing phenomenon and the Subject Sharing Hypothesis (Lee 1992).

The third part proposes an idea that SVCs can also be found in Japanese. In order to fulfil this task, general properties of SVCs are tested on Japanese and five different structures of Japanese V-V combinations are proposed based on verbal transitivity.

The fourth part of this thesis connects all the things already discussed and opens a new interpretation of previous approaches to Japanese V-V constructions, especially with respect to their terminology. A closer look is taken on differences between two terms - “compound verb” and “serial verb.”

The last part analysis English bare infinitives and compares them to already discussed Japanese and Korean SVCs.

In this thesis, plenty of examples from different languages are used in order to support my claims and demonstrate their validity. All of the examples taken from a different source are given together with the source of origin, which they were taken from. The remaining ones, especially from Japanese and Korean, are my own examples, after consulting with native speakers of both languages.

10 2 SERIAL VERB CONSTRUCTIONS: THE ENGLISH SUBCASE

2.1 What Are (NOT) Serial Verb Constructions

The notion of two verbs being serialized and functioning as one may for some speakers, especially for the speakers of those languages where such verb constructions are not common, represent structures difficult to comprehend or even difficult to get some notion of. On the other hand, there are definitely languages where the term SVCs might result in many suggestive examples of series of verbs, and therefore such speakers and their interpretations may vary. In this section, I introduce this topic within the confines of the English language. To successfully fulfil this task, I first comment on previous research in the field of “compound verbs” of English in Lieber (1983).

Let’s first illustrate the definition of SVCs made by Collins (1993, 91):

(1) Definition of SVC A serial verb construction is a succession of verbs and their complements (if any) with one subject and one tense value that are not separated by any overt marker of coordination or subordination.

According to this definition, the SVC is a sequence of verbs without any overt conjunction between them. This sequence has only one subject and only one tense value that are shared. From now on, I pick some sequences of English verbs to demonstrate what this definition might means for the concrete examples below.

(2) He is eating every hour.

(3) She has driven home.

The examples (2) and (3) are here to demonstrate what SVCs do not look like for a very simple reason - the SVC is a series, which contains two lexical verbs, whereas here an auxiliary verb is/has has been used together with a lexical verb. Furthermore, there are also markers of subordination - suffixex -ing/-en. Having

11 ruled out these two cases, let’s focus on examples when actually two lexical verbs are used in one sentence.

(4) I tried to ask a question.

Although the example (4) contains two lexical verbs, it is still different from constructions made by a serialization process. The crucial difference is in its structure that actually consists of one clause and one semi clause. When we speak about SVCs there is always only one clause with two verbs being serialized. Another significant difference is the presence of to infinitive marker. As in (1), there is no over marker between the series of verbs in SVCs.1

2.2 English Compound Verbs

Now, it seems that there can hardly be any combination of two English verbs that could resemble the structure of SVCs and satisfy all the conditions given by Collins in (1). However, there is one more structure that truly resembles the structure in serialization, and therefore it must be examined. Consider the examples taken from Lieber in (5).

(5) a. freeze-dry, drop-kick (Lieber 1983, 255) b. stir-fry, slam-dunk, roll-start (Lieber 1983, 265) c. blow-dry, spell-check (J. Emonds, pers. comm.)

The first thing to be noticed is that the examples given in Lieber (1983) are not very frequent. One may say that they are more like exceptions and therefore not many conclusions can be made. Yet, it seems that they truly satisfy all of the conditions given by Collins (1993) - one subject, one tense value, and no overt marker. However, there are further differences between the set of examples in (5) and the SVCs defined by Collins.

1 There are cases when an overt marker is inserted between two verbs in SVCs. It is only possible when this marker has no grammatical function. As is widely understood, the infinitive marker to has its grammatical function. 12 First, these structures in (5) are “true” compounds. It means that once they were formed, they function as a single unit - they function as one single verb. This also means that these already existing compounds cannot be interrupted by any syntactic or morphological elements. Therefore the application of the inflection process inside the compounds in (6) results in constructing ungrammatical sentences.

(6) a. *A player dropped-kick the ball. *He is stirring-fry the vegetable. b. A player drop-kicked the ball. He is stir-frying the vegetable.

The example (6) demonstrates how the process of inflection can be applied on compound verbs. As proposed earlier, verbs drop-kick and stir-fry are compounds. As for every compound, it always has at most one head and in the examples in (6), it is the second verb from the series, which works as a syntactic head due to the right-hand head rule (Lieber 1983). Therefore not the leftmost element, but the right one (the head) takes the inflection.

In order to support the idea that compound verbs are single units, the gapping operation is tested in this section. Gapping has been researched thoroughly in the work of Jackendoff (1971). According to Jackendoff, “the simplest case of Gapping deletes the verb of one or more clauses conjoined to the right of a clause containing the same verb” (1971, 21).

(7) Max ate the apples and Sally the hamburgers. (Jackendoff 1971, 21) (8) Max ate the apples and Sally (ate) the hamburgers.

The most elementary case of Gapping is illustrated in (7) where according to Jackendoff’s definition the second of the verbs can be deleted while the interpretation remains same (8).

(9) a. I stir-fried the vegetables, and my friends the mushrooms.

13 b. *I stir-fried the vegetables, and my friends stir the mushrooms.

The example (9) is to demonstrate the process of gapping applied to the compound verbs. As for the result of this operation, it seems that compound verbs behave as single units and they can only undergo the process of gapping when both of the verbs are gapped. When only one of the verbs is gapped, ungrammatical sentences are constructed.

Another way how to support the idea that compound verbs work as single units, can be demonstrated by assignment of stress. The field of stress assignment has been already analysed by Chomsky and Halle in the work called The Sound Pattern (1968). In this exhaustive work, two rules that seem to be crucial for the distinction between the compound and the syntactic structure have been introduced. First, Nuclear Stress Rule says, that primary stress is assigned “…to the second sonority peak” (1968, 92). Therefore the syntactic structures below in (10) have stresses on the second element.

(10) a. blue ‘sky b. white ‘board c. black ‘shadow

On the other hand, the Compound Rule proposes that primary stress is assigned “…to the first of the two sonority peaks” (1968, 92). According to this rule, every English compound should have its stress on the leftmost element.2

(11) a. ‘black board b. ‘green house c. ‘black bird

2 There are also some exceptions that have been noticed by Bloomfield. For instance, the word ice-cream can have either stress on the word ice or on the word cream, depending on if it is understood as a compound or as a syntactic unit (1933, 228). 14

According to the Compound Rule proposed by Chomsky and Halle, the primary stress falls on the leftmost element of the compounds. All of the examples in (11) validate this rule.

(12) a. ‘stir-fry b. ‘spell-check c. ‘freeze-dry

Having examined the stress assignment in the N-N compounds, stress and its placement must be examined on the V-V compounds. The examples in (12) behave as true compounds even with respect to stress assignment. The stress always comes to the leftmost of the items, and it is never placed on the second verb in a compound.

Furthermore, all the examples in (5) are only formed when two transitive verbs are used. 3 Lieber says that these two transitive verbs “…can somehow satisfy their argument structure with the same noun” (1983, 265). It seems that even though these compound verbs consist of two verbs representing two activities when used separately, they function as a single activity when they are used as a V-V compound.

Even though compound verbs to some extent satisfy Collins’ definition concerning SVCs, they are undoubtedly different. To support this claim, five different arguments have been given:

• No inflection in the middle of a compound. • Gapping shows that V-V is a single unit. • Stress shows a word-word compound that is X° not XP.

3 On the contrary, in languages with SVCs, transitivity does not matter, and combinations of either transitive or intransitive verbs are possible. To see some examples of serialized verbs with different transitivity, consult chapter 5, where syntax and transitivity of serial verb constructions are discussed. 15 • Both verbs are transitive. • There is only one single activity for a V-V compound.

To sum up, it has been demonstrated that compound verbs are single units. The structure of V-V in compound verbs is different from the structure of V-V in SVCs.

2.3 English Bare Infinitives

Besides the data on English compound verbs found in Lieber, which actually include only Transitive V2 and Transitive V1 pairs, I demonstrate that there are English V-V combinations of another different type.4

(13) a. come live with him b. help cook rice c. go eat sushi d. run tell him the story

The examples in (13) are to illustrate the presence of another type of a V-V combination. All of them have the same structure - the structure where the V2 is immediately followed by V1 in a bare infinitive. In contrast with the examples taken from Lieber in (5), neither of the verbs has to be transitive; V1 can be very productive; and lastly, they are not compounds. However, it is very interesting to observe how these V-V couples interact with inflection.

(14) a. She wants to go live with him. b. *She goes live with him. c. *She went live with him.

4 The terms V1 and V2 are not based on the order of verbs that appear in the syntax of languages with SVO word order. By V2, the verb, which takes inflection, is understood. V1 is the remaining one from the couple, which never takes inflection. 16 d. *She go lived with him.

(15) a. He will help cook rice. b. *He helps cook rice. c. *He helped cook rice. d. *He help cooked rice.

(16) a. I run tell Mary what was going on.5 b. We run tell Mary what is going on. c. *He runs tell Mary what is going on. d. He would run tell Mary what was going on. e. *He ran told Mary what is going on.

The examples in (14) - (16) show the restriction on V2 in terms of inflection. It seems that there cannot be any inflection on either V1 or V2 from the series. This means that the insertion of the 3rd person singular suffix -s results in ungrammatical structures. Similarly, the same thing happens when the past tense morpheme is added to the V1 or V2. Furthermore, the examples are also different from compound verbs, because the past tense morpheme cannot be attached to the V1, as well.

On the other hand, it is possible to use these V-V couples for 1st and 2nd person singular, plural, or after an auxiliary or a modal verb. This peculiar restriction does not contradict any of the conditions given in Collins (1993) - only one subject is shared; there is at most one head, which takes all the inflection (therefore there is only one tense value); and the construction is not separated by any conjunction or any overt marker.

Furthermore, even though this case seems to be restricted to only specific types of verbs appearing in the V2 position, such as go, come, help, run, they cannot be ruled

5 Note that the expression run tell is only colloquial (J. Emonds, pers. comm.). 17 out for they satisfy all the condition given by Collins.6 In other words, English does have a restricted type of serial verbs that has not been studied or even recognized in the literature yet.

2.4 Conclusion: English V-V and SVCs

To sum up, there are some examples in English that satisfy the criteria given by Collins. Five different arguments have been given in order to demonstrate that compound verbs differ from SVCs. Besides compound verbs, where two transitive verbs are coordinated, there are different ones, which may vary in their transitivity.

It is interesting to notice that even though the combination of lexical categories such as N-N and V-N are very abundant and productive in English (Lieber 1983, 255), the combination of two verbs (V-V) in the bare infinitive is not so frequent and there are actually only dozens examples of such constructions. However, this does not mean that they cannot be newly formed. It only seems that there is a limited set of verbs that may appear in the position of V2.

The objective of this section has been to propose an overview or a concept of how the V-V constructions and their interpretations may differ. This has been done from the background of English. In the next section, in order to get deeper in the phenomenon of SVCs, we will leave English for a while, and switch to a different language, which has plenty of such constructions.

6 The fact that only a certain type of verb may appear in the position of V1 is also common for other languages having SVCs. There is a closer look on this fact in later sections when discussing SVCs in Korean and Japanese. 18 3 GENERAL PROPERTIES OF SERIAL VERB CONSTRUCTIONS

3.1 Introduction

In this section, I start with more detailed analysis of SVCs especially in terms of their morphological and syntactic properties. So far there have been many works done on the phenomenon of SVCs appearing among many languages of Africa and Asia (Baker 1989, Schachter 1974). In those languages, SVCs occur mostly in languages with the Subject-Verb-Object (SVO) word order. However, for the later comparison with Japanese and its so-called “V-V compounds,” I will provide more detailed analysis of languages with a word order that is identical to the one in Japanese. Korean and Ijo are considered to be such languages.

The presence of SVCs in Ijo and Korean, has been already demonstrated by other linguists (Lee 1992; Williams 1963). The behaviour of serialization among Korean verbs has been analysed by a Korean linguist Sookhe Lee (1992). She is apparently the first linguist to start with syntactic and semantic analysis of SVCs in Korean and basically the first one to notice their similarities with their equivalents found in especially Bantu languages. Furthermore, she rules out Baker’s rule for the “object sharing” phenomenon, which I focus on in a later section (1989).

In the first place, I will pay attention to the definition of SVCs made by Collins (1993, 91), repeated from section 2.1 earlier. The definition is as follows:

(1) Definition of SVC A serial verb construction is a succession of verbs and their complements (if any) with one subject and one tense value that are not separated by any overt marker of coordination or subordination.

At this point, I have to state that the above-mentioned definition corresponds to properties observed in the majority of other linguistic works written on the

19 phenomenon of SVCs. Collins’ definition is brief, nicely formulated, and depicts the main characteristic of SVCs which are: succession of verbs, one subject, one tense, and no overt marker. In section 3.2, I start with the analysis of the last property of SVCs - succession of verbs without any overt marker, the property of English serial verbs discussed earlier in section 2.3.

3.2 No Overt Marker

In Korean, SVCs consist of two verbs mediated by the so-called ‘infinitive’ suffix ‘-e’ (Kang 1997, 6). A typical example of V-V serialization found in Korean is (2).

(2) Chelswu-ka cang’aymul-ul ttwi-e neom-eoss-da. Chelswu-Nom obstacles-ACC get on-L pass through-Past-Dec ‘Chelswu got through the obstacles’

One may wonder if the Korean example in (2) violates the definition in (1) stated by Collins because there is a lexical, suffix, or something that could suggest some kind of coordination between V1 and V2. Many other linguists have noticed and examined this troublesome morpheme. Lee calls this suffix a ‘dummy morpheme’ and successfully demonstrates that it has no meaning or any kind of grammatical function (1992, 123-124). Therefore it cannot be considered as any kind of conjunction or other marker that could rule out the existence of SVCs in Korean. She proposes that it is one of the properties or “deformation” of Korean verbs to appear necessarily in some kind of environment with at least one suffix. This property had already been noticed by Kang, when speaking about Morphological closure (1988, 78):

(3) Morphological closure Bound predicates must be closed off by a set of suffixes belonging to the category C.

Lee also mentions the definition in (3) and agrees on the fact that “…the category C consists of verb-final suffixes and does not include the aspectual/Agr/modal

20 suffixes in Korean”(1992, 123). Basically, this means that there is always a suffix attached to a Korean verb, though it does not necessarily have to carry any meaning or grammatical function. A suffix “-e” is one of the examples having neither grammatical function, nor meaning. More examples of such morphemes functioning only as a linker can be found in Ijo7.

3.3 Subject Sharing

A SVC is a series of verbs sharing only and just one single subject. It means that a sequence of two verbs having different subjects is not considered to be SVCs. An example of a true SVC where two verbs share one subject is below in the example (4).

(4) Chelswu-nun pam-ul kuw-e mek-ess-ta. Korean (Kang 1997,1) Chelswu-Top chestnut-Acc broil-Inf eat-Past-Dec ‘Chelswu ate the chestnut by broiling (it)’

Lee provides a set of examples from other languages representing structures with two subjects resulting in the creation of structures different from of SVCs. One of them is taken from Schachter (1974) and also mentioned in Lee (1992) is in (5).

(5) *Kofi yεε adwuma Kwaku maa Amma. Akan (Schachter 1974) Kofi did work Kwaku gave Amma.

Constructions with more than two verbs having two different subjects do not represent SVCs.

The second rule that is common for subject in SVCs is that the subject must be followed by V1 and V2. Structures when the syntactic subject appears between V1 and V2 are ungrammatical (Lee 1992, 19).

7 A more detailed analysis can be found in Williamson 1963. 21 3.4 Tense Sharing

Another syntactic property that has been noticed by Collins is that SVCs have only “one tense value”. Lee claims that besides tense, there are aspect and mood shared as well. She provides an example from Akan taken from Schachter (1974) where only one aspect item can be involved.

(6) a. Mekɔɔe mebaae. Akan (Schachter, 1974) I-went I-came ‘I went and came back.’

b. *Mekɔɔe maba. I-went I-have-come ‘I went and I have come back.’

Examples when only the aspect item is involved are grammatical (6a), when more than one marker of such a type (a combination of progressive aspect + non- progressive aspect, etc.) is used, it results in ungrammatical sentences (6b) (1992, 20-22).

3.5 Negation

Negation is one of the properties that is common to SVCs but has not been mentioned in Collins “Definition of SVCs” in (1). According to Lee, the Negative Polarity Items (NPIs) “are constructed with each other under strict clause- boundedness”(1992, 130). In Korean, there are only a few NPIs - “amwu-do” (meaning nobody) and a prefix an (meaning simply “no”/”negation of the action”). Lee provides a set of examples (7) displaying the behaviour of NPIs where “amwu- do” and “-an” must appear in the same clause in order to form grammatical constructions (1992, 131):

(7) a. Chelswu-nun [amwu-to kecismal-ul an ha-ess-ta-ko] sayngkakha-n-ta. -Top[nobody lie-Acc Neg do-Past-Dec-Comp] think-Pres-Dec ‘Chelswu thinks that nobody told a lie.’

22 b. Amwu-to [Chelswu-ka kecismal-ul ha-ess-ta-ko] sayngkakha-ci anh-nun-ta. Nobody [ -Nom lie-Acc do-Past-Dec-Comp] think-CI do not-Pres-Dec ‘Nobody thinks that Chelswu told a lie.’

If NPIs with negative polarity appear outside of the clause, the structures will be judged ungrammatical:

(8) a. *Chelswu-nun [amwu-to kecismal-ul ha-ess-ta-ko] sayngkakha-ci anh-nun-ta. -Top[nobody lie-Acc do-Past-Dec-Comp] think-CI do not-Pres-Dec ‘Chelswu thinks that nobody told a lie’

b. *Amwu-to [Chelswu-ka kecismal-ul an ha-ess-ta-ko] sayngkakha-n-ta.. Nobody [-Nom lie-Acc Neg do-Past-Dec-Comp] think-Pres-Dec ‘Nobody thinks that Chelswu told a lie.’

However, SVCs behave differently and therefore are presumably ungrammatical structures, when the verb with NPI “-anh” appears outside of the clause with the other NPI such as “amwukes-to” or “amwu-do”, result in forming grammatical constructions (Lee 1992, 132). Examples of such cases are in (9).

(9) a. Chelswu-nun [amwukes-to mantul-e] mek-ci anh-ass-ta. -Top [nothing make-L] eat-CI do not-Past-Dec ‘Chelswu did not cook nor eat anything.’

b. Amwu-to [pizza-lul an mantul-e] mek-ess-ta. Nobody [pizza-Acc Neg make-L] eat-Past-Dec ‘Nobody made and ate pizza.’

Examples in (9) demonstrate tense sharing in SVCs. In other words, polarity scope has been extended over two clauses, and therefore the whole sentence has negative polarity. If the negative polarity had scope only over one clause, the structure would be ungrammatical.

3.6 Headedness

SVCs are generally constructions when a sequence of two or more verbs with only one verb behaving as a head of the whole construction. This means that either V1

23 or V2 takes verbal inflection. Korean is a SOV language, and therefore its head is on the right (V2).

(10) a. John-i kaymi-lul nul-e cuki-ess-ta. Korean (Kang 1997, 17) -Nom ant-Acc crush-L kill-Past-Dec ‘John killed the ant by crushing it.’

b. *John-i kaymi-lul nul-ess-e cuki-tta. -Nom ant-Acc crush-Past-L kill-Dec ‘John killed the ant by crushing it.’

Any kind of inflection on the V1, for example a past tense suffix -as in (10b) being attached to the V1 in the sequence in Korean, will result in formation of ungrammatical sentences.

The most prominent properties of SVCs are: no overt marker, subject sharing, tense sharing, negation, and headedness. Furthermore, it seems that there is only one “single interpreted event” in SVCs and therefore the structure of the SVC is also different from the structure in (11) where two events are represented.

(11) She decided yesterday to leave today.

In this section, I have illustrated the typical properties of SVCs among various languages. Due to the similarities with Japanese, my focus has been especially on Korean. In the following chapter, I will demonstrate an idea that SVCs are also represented in Japanese.

24 4 JAPANESE SERIAL VERB CONSTRUCTIONS

4.1 Introduction

In section 3, I have introduced the main properties of SVCs. These properties were previously examined by the most influential linguists in the particular area, such as Collins, Lee, and Baker. Consequently, all of the definitions were provided with suitable examples taken mostly from the African languages and Korean. In this chapter, I am focusing on another language - Japanese. I propose that SVCs are also to be found in sequences of Japanese verbs. After I illustrate the presence of SVCs among Japanese structures, I will investigate their syntactic properties that I consider to be identical with their Korean equivalents.

To begin with, I give a couple of examples of V-V serialization in Japanese. The examples in (1) are included in order to show how such a hypothetical SVC in Japanese may look.8

(1) a. Ten’in-ga Nao-ni ayamari-taosi-ta. waiter-Nom Nao-Dat apologize-defeat-Past ‘Waiter kept apologizing to Nao until she forgave him.’

b. Obaatyan-ga itumo uwasa-o ii-tirasu. granny-Nom always rumour-Acc say-spread ‘Granny always spreads the rumour.’

There is a sequence of two verbs sharing an argument in both sentences above. It is a “waiter”, who is Agent of both, “apologizing” and “defeating” in (1); it is the “granny”, who is the Agent of both, “saying” and “spreading” in (1)9. At this point, one may notice the similarity between SVCs and this type of Japanese sequence of

8 Although the interpretations of following sentences are already included I want the reader to know that other interpretations are also possible. However, there should always be one interpretation that is more prevailing than the others. 9 There is another argument (an object uwasa) for the sequence of verbs in (1) being shared by the two verbs due to their θ-roles (both verbs are transitive). 25 verbs. This whole section attempts to support this idea and make clear whether Japanese also has so-called SVCs or it has a sequence of verbs of some other type.

4.2 No Overt Marker

The examples in (1) represent typical V-V constructions in Japanese when two verbs are connected by a so-called ren’yoo form. This form is sometimes called an ‘i-form’ because mostly it truly ends with the letter with the sound of /i/. The structure, which looks like - [V1-i V2], caries no meaning and no grammatical function, and its existence is due to the property of Japanese verbs to inevitably end in a suffix.10 The examples in (2) demonstrate different usage of the ren’yoo form.

(2) a. Taro-wa eki-e iki, kaimono-o si-ta. Taro-Top station-Gen go shopping-Acc do-Past ‘Taro went to the station and did shopping.’

b. Taro-wa eki-e iki, kaimono-o suru. Taro-Top station-Get go shopping-Acc do-Pres ‘Taro will go to the station and do shopping.’

c. Taro-wa eki-e iki, kaimono-o suru. Taro-Top station-Get go shopping-Acc do-Pres *’Taro went to the station and he will do the shopping.’

There are three constructions created by using so-called ren’yoo form in examples (2).11 We can observe that there is a sharing of tenses between the two verbs in two clauses. In other words, it means that ren’yoo form does not carry any grammatical function such as tense because the meaning of it in (2a) and (2b) would have to be different due to the tense morpheme attached to the verb in the second clause. Therefore structures with different interpretations like (2) that

10 Notice that the same property/deformity has already been introduced in section 3.2 in the example (3) w.w. “Morphological closure” in Korean. 11 Notice that even though we are not speaking about ren’yoo form inside two verbs but rather as a way of connecting two clauses together, the properties of such a form remain identical. 26 would suggest some independence in grammatical function of the ren’yoo form turn out to be ungrammatical.

Another case when ren’yoo form is used is when we address someone by using honorific language. There are many ways how to address someone in a polite way in Japanese depending on each situation. However, there is a general way to do so when we speak about actions that the superior person is doing. In these cases, a honorific prefix -o is firstly attached to the verb, which is in ren’yoo form. Secondly, a Dative particle -ni comes after the verb. Lastly, the verb naru, meaning ‘to become’, is attached to form correct honorific expressions. 12 Two of such cases are in the example (3).

(3) a. Sensei-ga ronbun-o o-kaki-ni nat-ta. Teacher-Nom thesis-Acc Hon-write-ni become-Past ‘The teacher wrote a thesis.’

b. Sensei-ga ronbun-o o-kaki-ni nar-u. Teacher-Nom thesis-Acc Hon-write-ni become-Pres ‘The teacher will write a thesis.’

In examples (3), it is essential to notice the fact that the tense morpheme can only be attached to the naru verb (meaning ‘to become’). Although the ren’yoo form remains the same in both examples above, the interpretation of Tense is different - past tense in (3) and present tense in (3). It supports the idea that ren’yoo form does not carry any grammatical function (e.g. Tense).

The structure of ren’yoo form [V1-i V2] is very similar to that in Korean where we have also structure like [V1-e V2]. As mentioned in previous sections, Lee calls -e a “dummy morpheme” and suggests that it has no semantic or syntactic meaning

12 One may wonder why the verb naru meaning ‘to become’ is employed as a means to construct honorific expressions. This concept arises from the linguistically fossilized attitude of Japanese people and their inability to address the superior directly when speaking about superior’s actions. Therefore, instead of saying ‘he did something’, they basically prefer to say that ‘he became doing something,’ (J. Emonds, pers. comm.). 27 (1992, 122). I have similarly demonstrated that Japanese ren’yoo form has no meaning or any grammatical function.

4.3 Subject Sharing

There have many discussions about argument sharing in the previous works concerning SVCs (Baker 1989, 516; Collins 1993, 93). What is essential is the fact, that both Baker and Collins have suggested that object sharing (internal argument) is obligatory. Collins (1993, 93) provides a definition for “Argument Sharing in SVCs”:

(4) In a serial verb construction, V1 and V2 must share an internal argument.

Both of them provide many examples taken from Ewe (Collins 1993), Yoruba and the Sranan language (Baker 1989).

(5) me’a nu ‘u Ewe (Collins 1993, 34) I cooked thing ate ‘I cooked something and ate it’

The example (5) represents a typical V-V construction where two transitive verbs share one internal argument (nu - ‘thing’). The same type of argument sharing occurs in Korean. Focus on the example (6), which corresponds with Collins’ and Baker’s theory about argument sharing, where V1 and V2 (while being connected by a ‘dummy morpheme’/linker) share the internal argument (object) - ‘laundry’.

(6) Nay-ka ppal lay-lul nel-e noh-ass-ta. Korean I-Nom laundry-Acc spread-L put-Past-Dec ‘I hang the laundry.’

Meanwhile, Lee introduces a Subject Sharing Hypothesis (SSH) that actually claims the theory about internal argument sharing to be incorrect, or rather, not always applicable (1992, 75). The SSH is stated in (7).

(7) Subject Sharing Hypothesis (SSH) a. Only if both verbs are transitive must there be an object sharing; b. all and only serial verbs share a subject. 28 Moreover, Lee comments on the SSH: “The SSH predicts that every possible combination of two verbs based on their transitivity occurs in SVCs. If both verbs are transitive, they must share an object as well as subject”(Lee, 1992, 76). This is an innovative idea in terms of SVCs because it rules out Baker’s concept of object sharing. For example, Lee also claims that the serialization of two verbs that are intransitive is also possible.13 This hypothesis is supported by the following examples:

(8) a. Pihayngki-ka (hanul-lo) nal-a olun-ta. Korean airplane-Nom (sky-Dat) fly-L go.up-Det ‘A plane takes off (into the sky).’

b. Watashi-ga naki-tukare-ta. Japanese I-Nom cry-tire-Past ‘I got tired out of crying’

The examples (8) illustrate how two intransitive verbs in serialization share their external argument - Subject. Furthermore, there is only one linker, one activity, and one tense. This Subject is only one and due to their intransitivity, there is no reason for them to share any other argument. As one can see, these types of constructions can be found in both Korean and Japanese. Moreover, these cases are very common and used on a daily basis. This discussion concerning θ-roles (semantic roles) and argument structures will be discussed in detail in later chapters.

4.4 Tense Sharing

The issue about tense sharing has already been discussed in section 4.2 when speaking about “No Overt Marker” rule. In order to avoid duplication, I just briefly sum up the basic concept without providing any new sets of examples.

13 Bear in mind that the sequence of two transitive verbs must share an object (Lee 1992, 109). 29 In Japanese V-V constructions, V1 is used in so called ren’yoo form, which does not carry any meaning or grammatical function. It means that the tense cannot be interpreted only by looking at the V1 in the ren’yoo form, but the existence of V2 with its suffixes (agglutinating suffixes where each of them carry one and only one grammatical function) is necessary to express tense, modality, mood, or level of politeness. In other words, features such as tense, modality, etc. are shared by both of the verbs. To review, compare examples in section 3.2 with examples (2).

One may wonder why it is only one of the verbs or wonder why it is V2 that takes the whole inflection. I will answer these questions in section 4.6, concerning the headedness of SVCs.

4.5 Negation

Negation has been previously noticed as a shared feature in SVCs by many linguists. I have introduced this topic in section 3.5, especially using the background of the Korean SVCs. Due to the similar behaviour of Japanese verbs I will use the same approach and linguistic tests. Consider the following examples:

(9) a. Momoko-wa [dare-mo uso-o tuite-ina-i to] omo-tta. -Top [nobody lie-Acc attach-Neg-Pres-Comp] think-Past ‘Momoko thought that nobody told a lie’

b. Dare-mo [momoko-ga uso-o tui-ta to] omowa-na-i. Nobody [ -Nom lie-Acc attach-Past-Comp] think-Neg-Pres ‘Nobody thinks Momoko told a lie.’

There are basically two ways (two types of negative polarity items (NPIs)) to express negative polarity in Japanese. One of them is a -nai suffix “not” that can be attached to the verb stem and results in making negative structures. The other way to express polarity is to use a “-mo” suffix, which can be attached to the nouns (e.g. daremo - ‘nobody’, hitorimo - ‘not a single person’, Momokomo - ‘not even Momoko’). All of these NPIs must appear inside of the negative scope, otherwise structures will be ungrammatical. The example (9) demonstrates the grammatical

30 usage of NPIs “-nai” and “-mo.” As one can see, they appear in the same clause due their negative polarity, whereas these examples cannot appear in a clause which does not have negative polarity.

(10) a. *Momoko-wa [dare-mo uso-o tui-ta to] omowa-na-i. -Top [nobody lie-Acc attach-Past-Comp think-Neg-Pres ‘Momoko thought that nobody told a lie’

b. *Dare-mo [Momoko-ga uso-o tuite-nai to] omot-ta. Nobody [ -Nom lie-Acc attach-Past-Comp] think-Past ‘Nobody thinks Momoko told a lie.’

Examples in (10) show the ungrammaticality of sentences where NPIs appear outside of the negative polarity scope.

(11) Momoko-wa [nani-mo tukuri] hajimete-ina-katta. -Top [nothing cook] start-Neg-Past ‘Momoko did not start to cook anything.’

When we focus on the polarity scope inside of the structures including serialization, we notice that the polarity is somehow extended. We realize that the polarity is shared by both verbs in serialization. In the example (11), there is a -nai suffix attached only to the V2 “hajime-ru” but the whole structure has a negative interpretation. Negative scope is shifted and therefore the -mo NPI can be attached even in seemingly positive structure. However, the negative interpretation of the sentence and the presence of NPI indicate the negative structure.

To sum up, negative polarity is extended over the whole SVC. This fact validates the claim that verbs in SVCs function as “one”.

4.6 Headedness

A SVC is a sequence of verbs where only one of them works as a morphological and syntactic head. Firstly, this means that only the head takes any inflection. Secondly, the head can also influence other elements in the structure. In languages with SOV

31 word order, it is usually V2 that takes the role of a syntactic head. In this section, I concentrate on the behaviour of Japanese SVCs with respect to their headedness.

(12) a. Watasi-ga kaki dasi-te-i-na-katta. Japanese I-Nom write take.out-TE-be-Neg-Past ‘I did not start to write.’

b. *Watasi-ga kaka-nai dasu. I-Nom write-Neg take.out ‘I did not start to write.’

The examples (12) demonstrate the fact that only the head (V2) takes inflection. If there is any kind of inflection (e.g. Negative Polarity Item) attached to the other element (V1), the structures will be ungrammatical (12).

Previously, there has been an interesting approach concerning this field by Nishiyama. He is a Japanese linguist who has noticed some similarities of Japanese V-V constructions with so called SVCs found in other languages. I will comment on his work in later sections. For the moment, I only focus on his work concerning headedness. Examples taken from his work can be seen in (13) (Nishiyama 1998, 177).

(13) a. John-ga Mary-o ot-ta. -Nom -Acc chase-Past ‘John chased Mary.’

b. John-ga Bill-ni tui-ta. -Nom -Dat attach-Past ‘John attached to Bill.’

There are two verbs in examples (13). The verb “ot-ta” (the past tense of “o-u”) in the former example is an accusative transitive verb so the Japanese particle “o” must be attached to its accusative object in order to form grammatical structures. The particle “o” is attached to the object Mary, and therefore the structure is grammatical. However, a different type of construction is used in the later example

32 (13). According to Nishiyama, “…tuk ‘attach’ takes a dative complement…” (1998, 177), which is expressed by a different Japanese particle, the dative particle - “ni”, that is attached to “Bill”.

However, there are cases when these two verbs appear in serialization (oi-tuk) Therefore, one may wonder how such a structure would look and ask questions if the verb takes an accusative object or dative complement. See the structure in (14) made by Nishiyama (1998, 177):

(14) a. John-ga Mary-ni oi-tui-ta. -Nom -Dat chase-attach-Past ‘John chased and attached to Mary.’

b. *John-ga Mary-o oi-tui-ta. -Nom -Acc chase-attach-Past ‘John chased and attached to Mary.’

Nishiyama understands this as not a kind of restriction of Case, but rather some restriction concerning complement structure. He claims that Japanese V-V constructions cannot have different complements and therefore the case of the V2, which is a head of the construction, is assigned in (14) to form grammatical structures (Nishiyama 1998, 177-184). Structures when the case of the V1, instead of V2’s case, is assigned, will result in constructing ungrammatical sentences (14). Some other examples when a V-V construction takes only one NP due to the right- headedness of such a construction, are as follows:

(15) a. sigoto-ga si-agat-ta Japanese (Nishiyama 1998, 189) job-Nom do-get.ready-Past ‘The job was done and got ready’

b. coat-ga ki-kuzure-ta Japanese (Nishiyama 1998, 189) -Nom wear-get.out.of.shape-Past ‘The coat was worn and got out of shape.’

33 The examples in (15) are included to support the idea of Nishiyama. The combination of either transitive or intransitive verbs in V-V is, they will work as one, with one primary head. Due to the SOV word order, usually, it is the V2 that functions as a structural head14.

4.7 Conclusion concerning SVCs in Japanese

In this chapter I have proposed an idea15 that SVCs, commonly known in languages of West Africa, in Korean, Indonesian, and many other languages, can be found also in the Japanese language. In order to establish this idea I have shown that various morphological and syntactic tests applicable in most languages with SVCs can be found even in the context of Japanese. To sum up, I have created Table 1, which is briefly summarizing that properties typical for SVCs can be also found in Japanese.

(16)

Properties of SVCs SVCs in Other SVCs in Languages Japanese

No Overt Marker yes yes

Subject Sharing yes yes

Tense/Aspect/Mood yes yes Sharing

Shared yes yes Negation/Polarity

Headedness (One yes yes syntactic head in SVC)

14 At this point, it must be said that there are exceptions and cases when not V2 but V1 functions as a head at least in terms of transitivity. A closer look at this phenomenon will be taken in a later chapter when speaking about combinations of verbs in V-V structure and their transitivity. 15 This idea was previously proposed by Nishiyama (1998). However, I found it insufficient for the incomplete list of arguments defending this theory and the lack of the conclusion concerning this phenomenon. In other words, despite giving birth to a new idea concerning SVCs in Japanese, Nishiyama still continues to use very traditional approach and keeps using out dated terminology -‘compound verbs’. 34

Single interpreted event yes yes

Table 1: Properties of SVCs - Japanese

35 5 SYNTAX AND TRANSITIVITY OF SERIAL VERB CONSTRUCTIONS

5.1 Argument Structure

In the last decades, there have been several discussions concerning argument structures of SVCs. There are those which support the idea that either object or NP must be (if possible) shared by both V1 and V2 even though the second of the verbs might be an intransitive one (Baker 1989). On the other side, there are those that are against the obligatory sharing of the internal NP (object) by both of the verbs and therefore have come up with a new type of analysis. According to Lee, “Once the θ-Criterion is satisfied by the θ-marking of V1, the V2 is not required to obligatorily θ-mark the object of the V1 unless the V2 also obligatorily θ-marks that NP.” (1992, 85). This note concerning θ-roles is crucial for defining SVCs because some kinds of SVCs found in other languages would be ruled out from this category if the internal argument-sharing hypothesis was compulsory.

According to the Subject Sharing Hypothesis proposed by Lee (1992), it seems that there are many possible combinations of argument structures of verbs in SVCs. In this section, I provide a set of 5 possible argument structures found in SVCs that have been previously demonstrated in Korean (Lee 1992, 91-103). Consequently I extend this analysis to Japanese and provide its possible argument structure combinations, which seem to be very similar to the ones found in Korean.

Based on this definition proposed by Lee, 5 types of SVCs are possible.16

16 Recall that the head final languages such as Korean and Japanese will have different structures than languages with SVO word order. There is a closer look on the topic of headedness of Korean structures in chapters 3.6, and of Japanese structures in 4.6. 36 5.2 Intransitive V1 + Intransitive V2

The first group of the possible verb combinations is a combination of two intransitive verbs sharing their external argument. This combination (together with the combination of Transitive V1 + Transitive V2) prevails, compared to other remaining combinations. There are many examples of Intransitive V1 + Intransitive V2 combination found in Korean in (1), and similarly in Japanese in (2).

(1) a. Ku-nun hunukki-e wul-ess-ta. Korean (Lee 1992, 102) he-Top sob-L cry-Past-Dec ‘He sobbed.’

b. Pihangki-ka nal-a olun-ta. plane-Nom fly-L go.up-Dec ‘The plane took off.’

(2) a. Papa-ga warai tukare-ta. Japanese Father-Nom laugh tire-Past ‘Father got tired out laughing.’

b. Seito-ga tachi agat-ta. student-Nom stand rise-Past ‘The student stood up.’

The examples (1) and (2) include V-V combination of two Intransitive verbs being serialized by either the Korean ‘dummy morpheme’ -e or the Japanese ren’yoo form. The structure of (2), which is identical in all examples in (1) and (2), is illustrated in (3).

37 (3) S

NP V’

Det V’ V2

Papa V1 tukareru

warai

The tree structure in (3) represents Intransitive V1 + Intransitive V2 combination of verbs in SVC, which, as illustrated previously, does not include any internal argument sharing. The only θ-role is attached to the external argument (Subject). In other words, no Object is part of this construction. All of the properties demonstrated in sections 3 and 4 (tense sharing, headedness, no overt marker, negation) also work for the examples (1) and (2), and consequently does not violate their tree structure validated in (3).

5.3 Intransitive V1 + Transitive V2

Having discussed the SSH saying that the object sharing is not necessary for SVCs, structures consisting of Intransitive V1 + Transitive V2 are also considered to be grammatical and fit the frame work of SVCs. Lee provides a set of examples taken from Korean:

(4) a. Ku-nun Swuni-lul sorichi-e pwul-ess-ta. Korean (Lee 1992, 97) he-Top Swuni-Acc shout-L call.out-Past-Dec ‘He called out for Swuni.’

b. Ku-nun tam-ul ttway-e nem-ess-ta. he-Top wall-Acc jump-L go.over-Past-Dec ‘He jumped over the wall.’

38 There are two examples of Intransitive V1 + Transitive V2 serialization in (4). However it seems that V1 is a transitive verb due to its position, when it immediately follows the Object, but both V1s in (4) are unergative intransitive verbs. Therefore the structure V’, where transitive V2 assigns a θ-role directly to the NP is hierarchically higher than intransitive V1, which does not directly assign θ-role to the same NP, as illustrated in (6).

To support the idea that both verbs in the position of V2 in the example (4) are transitive even when they appear alone, see examples in (5). Recall the Accusative particle lul, which is used in both cases, is assigned to objects of the verbs.

(5) a. Ku-nun phili-lul pwul-ess-ta. He-Top flute-Acc blow-Past-Dec ‘He blew the flute.’

b. Ku-nun san-ul nem-ess-ta. He-Top mountain-Acc cross-Past-Dec ‘He crossed the mountain.’

Both examples in (4) have the structure identical with the tree structure in (6). The existence of the structure supports the validity of the SSH (Lee 1992, 96-97). 17

(6) (Lee 1992, 98)

The same structure with the combination of Intransitive V1 and Transitive V2 can be also found in Japanese. Examples of such structures are represented below:

17 I am aware of the fact that the flat structures violate the usual Minimalist assumption based on Kayne (1994), which assumes that all syntactic structures are binary. Yet, flat structures are used in this thesis as adaptations of these proposed by both Baker (1989) and Lee (1992) and as far as I can see, provide the closest match to the actual empirical paradigms of serial verbs cross linguistically. 39 (7) a. Kaisa-ga saito-o tati age-ta. Japanese company-Nom website-Acc stand start-Past ‘The company put up a website.’

b. Megumi-ga kangeki-de me-o naki harasi-ta. Megumi-Nom deep.emotion-DE eye-Acc cry cause.to.swell-Past ‘Megumi wept her eyes out.’

In (7), the transitive V2 ageru meaning ‘start’18 assigns a θ-role to the NP saito meaning ‘website’, which is suggested by the presence of the Accusative marker -o being attached to the NP (object) saito. The intransitive V1 however does not assign any direct θ-role to the NP saito. The intransitivity of V1s from the above examples can be observed when they are used alone, without V2s following them. Such cases are in the examples below:

(8) a. Seito-ga ikinari tat-ta. Japanese pupil-Nom suddenly stand-up-Past ‘The pupil suddenly stood up.’

b. Kinou kanozo-ga nai-te-i-ta yesterday she-Nom cry-TE-be-Past ‘She was crying yesterday.’

Back to the example (7), another thing to be noticed is that there is no object sharing, and verbs are connected by a shared subject, which again supports Lee’s SSH. It seems that the V2 is the head of the V-V construction. Therefore it takes all morphology (inflection) and also functions as a syntactic head (the presence of the Object and the Accusative marker is necessary due to the fact that V2 is transitive). However, it is V1, which carries the basic meaning of the V-V combination, and V2 only seems to modify or specify the whole action of V-V. The example (7) has the following structure:

18 Notice the fact that ageru belongs to the group of aspectual verbs, and therefore it loses its full meaning during serialization. Instead of its full meaning ‘raise’, its meaning is reduced to ‘start’ or rather to ‘begin the action’ when it is used in SVC in the position of V2. This disposition is also very common for other languages with SVCs. 40 (9) S

NP V’

kaisa NP V’ V2

saito V1 ageru

tati

Before we move to the next section I want to illustrate a seemingly different type of Intransitive V1 + Transitive V2 combination in Japanese, which is demonstrated in (10).

(10) a. Neko-ga asobi dashi-ta. Japanese cat-Nom play take.out-Past ‘The cat started to play.’

b. Nana-ga warai hazime-ta. Nana-Nom laugh start-Past ‘Nana started to laugh.’

The examples in (10) once again include V-V combinations, which consist of Intransitive V1 and Transitive V2. However in this case, the Transitive V2s do not have any internal arguments (Objects) that would satisfy their argument structure. Moreover, in cases when the Object is inserted in order to form a seemingly grammatical structure, ungrammatical ones result, as in (11).

(11) Neko-ga *ball-o ugoki dashi-ta. cat-Nom ball-Acc move take.out-Past ‘The cat started to play with the ball.’

The question is, how do these V2, which according to Nishiyama (1998) are syntactical and morphological heads, satisfy their argument structures? One may suggest that it is V1, not V2, which is the syntactic head due to the Nom particle -ga

41 attached to the Agent of the action. However, I consider such conclusions to be incorrect and demonstrate that the argument structure of V2 remains satisfied. It is not necessary for a transitive verb to have its internal argument. It should be the other way round - if there is an internal argument, it must have a verb, to be assigned to, not the opposite way19 . Furthermore, this absence of Object of Transitive V2 is very common especially when we discuss aspectual verbs in the position of V2. They modify the whole action, motion or movement, and therefore the presence of the internal argument is not necessary. This type of SVC in a way supports Nishiyama’s (1998) idea about primary headedness (morphological head), but in another way weakens his arguments about V2 being also a syntactic head.

5.4 Transitive V1 + Transitive V2

The combination of two transitive verbs is as frequent as with the combination of two intransitive ones. In this section I will illustrate such a case on some examples in Korean and similarly in Japanese.

(12) a. Na-nun sokum-ul ccik-e meke-ss-ta. Korean I-Top salt-Acc put.in-L eat-Past-Dec ‘I put in salt and ate it.’

b. Nay-ka pul-ul phiu-e palkhye-ss-ta. I-Nom fire-Acc light-L brighten-Past-Dec ‘I lighted up the room by fire.’

The examples from Korean in (12) represent the only case of obligatory internal argument sharing found among all SVCs. This is because of both verbs being transitive, which requires the presence of not only the Subject but also one more argument. For instance, in (12), the NP sokum, meaning ‘salt’ is shared by both verbs - ccikta, meaning ‘put in’, and mekta, meaning ‘eat’. The cases with identical structures also appear in Japanese. To compare, see the following examples:

19 This fact was also noticed by Nishiyama (1998) 42 (13) a. Ore-ga soturon-o kaki oe-na-katta. Japanese I-Nom final thesis-Acc write finish-Neg-Past ‘I didn’t finish writing the final thesis.’

b. Kanozyo-ga uwasa-o ii tirashi-ta girlfriend-Nom rumor-Acc say disperse-Past ‘The girlfriend kept spreading the rumor.’

The above examples represent structures identical to those in (12). Once again, besides the presence of Subject, it is necessary for the combination of Transitive V1 + Transitive V2, to additionally have Objects. In (13), the Subject Ore and the Object soturon are shared by a verb kaki, meaning ‘write’, and oeru, meaning ‘to finish’. This sentence has the following structure:

(14) S

NP V’

Ore NP V’ V2

soturon V1 oeru

kaki

To sum up, the above structure represents the case of Transitive V1 and Transitive V2 sharing both Object and Subject. This combination seems to be, together with the combination of two intransitive verbs, the most common. For this case, binary branching is not suitable because both V1 and V2 are transitive. Therefore both of them require internal argument (object) soturon.

43 5.5 Transitive V1 + Intransitive V2

The next representative of SVCs is a combination of Transitive V1 + Intransitive V2. It is very interesting to observe how this formation is represented in Korean:

(15) a. Ku-nun Mr. Kim-ul chac-a ka-ss-ta. Korean (Lee 1992, 99) he-Top Mr. Kim-Acc look.for-L go-Past-Dec ‘He went to see Mr. Kim.’

b. Ku-nun tongsayng-ul chac-a nase-ss-ta. he-Top brother-Acc look.for-L go.off-Past-Dec ‘He went to meet Mr. Kim.’

In the first place, let’s take the sentence (15) and try to analyse it. The subject ku, ‘he’ is shared by both verbs chacta ‘look for’ and kata ‘go’. The debate that might arise is if the Object of the Transitive V1 - Mr.Kim, is also shared by the intransitive V2. According to Lee’s SSH, the only case when the Object is shared is when two transitive verbs appear, and therefore in this case when V2 is intransitive, its Object is not shared. Furthermore, another argument for such a statement is that V2 semantically modifies the whole action of the Subject, which ‘is looking for Mr. Kim.’ Similar cases when Transitive V1 and Intransitive V2 appear to form a construction can be again found in Japanese and are depicted in (16).

(16)

a. Omise-ga tate kon-de-i-ta. Japanese shop-Nom stand.up be.crowded-Asp-be-Past ‘The store was crowded.’

b. Kesa watasi-ga ii make-ta. this morning I-Nom say lose-Past ‘I lost a quarrel this morning.’

Although the above examples represent the Transitive V1 + Intransitive V2 combination, both lack the presence of the Object and its Accusative marker o. When either verb tateru ‘to stand up’, or the verb iu ‘to say’ are used separately

44 without their V2 parts, the presence of the Accusative marker is obligatory, as in (17).

(17) a. Boku-ga keikaku-o tate-ta. Japanese I-Nom plan-Acc stand.up-Past ‘I made a plan.’

b. Obatan-wa zoudan-o i-wana-i. granny-Top joke-Acc say-NEG-Pres ‘Granny does not say any jokes.’

Furthermore, if the V2 is used in separation and the Accusative marker with internal argument is omitted, it results in ungrammatical sentences (or causes a meaning shift), as in (18).

(18) a. *Boku-ga keikaku-o tate-ta. b. *Obatan-wa zoudan-o i-wana-i.

This behavior, when the argument of an Accusative verb is not represented in Japanese V-V constructions, has been already noticed by Nishiyama (1998) and mentioned in this work in section 5.3. In order to avoid repetition, I will just briefly sum up this concept when V2 technically becomes both morphological and syntactic head of the whole construction. I assume that there is a tendency to drop or omit the internal argument of the V1 in the V-V construction. However, it does not mean that such cases do not exist. Notice the example below.

(19) Nana-ga gohan-o tabe owat-ta. Japanese Nana-Nom meal-Acc eat finish-Past ‘Nana finished eating the meal.’

45 (20) S

NP V’

Nana V’ V2

(NP) V1 owatta

gohan tabe

The example (19) has a tree structure illustrated in (20), which represents a structure for every Transitive V1 + Intransitive V2 combination. The transitive verb taberu ‘to eat’ requires the Object gohan ‘the food’. V2 - owaru ‘to finish’, is however intransitive and therefore Object sharing does not take place.

5.6 Transitive V1 + Ditransitive V2

This fifth section pays attention to the cases when ditransitive verbs are serialized20. The only possible combination is when the V1 is transitive, and the V2 is ditransitive. Three different ditransitive verbs are used to form SVCs in the example (21).

(21) a. Hirosi-ga tegami-o kazoku-ni kaki okut-ta. Japanese Hirosi-Nom letter-Acc family-Dat write send-Past ‘Hirosi wrote and sent letter to his family.’

b. Watanabe-ga bara-o tuma-ni kai atae-ta Watanabe-Nom rose-Acc wife-Dat buy give-Past ‘Watanabe bought rose as a gift for his wife.’

c. Mei-wa yatin-o oyasan-ni harai watasi-ta. Mei-Top rent-Acc land.lord-Dat pay carry-Past1

20 It is interesting to notice that this combination of Transitive V1 and Ditransitive V2 is not listed in Lee (1992) and therefore it opens a discussion whether Korean lacks or has this type of construction, as well. 46 ‘Mei forgot to pay her landlord the rent.’

Before I illustrate the tree structure of Transitive V1 + Ditransitive V2 combination, I briefly depict what the structure of the sentence in (21) and its constituents look like. As previously, a series of verbs shares one single Subject Hiroshi. Secondly, there is an internal argument shared by both verbs - the Object tegami ‘letter’. However, there is one more internal argument kazoku ‘family’ which is assigned to only one of the verbs from the series - V2. Due to its ditransitivity, besides Agent and Patient, the role of Recipient is also crucial in order to form grammatical structures.

(22) S

NP V’

Hirosi V’ NP V2

tuma ataeta

NP V1

bara kai

The structure of SVCs when Transitive V1 and Ditransitive V2 is serialized is exhibited in (22). The object bara is shared by both V1 kai and V2 ataeru. Due to the ditransitivity of the V2 only, there is one more object tuma, which is shared only by V2. For this reason, this structure cannot be represented by binary branching.

47 5.7 Concluding Remarks

In this chapter, I have illustrated 5 different structures that had been previously proposed by Lee (1992). Furthermore I have extended this approach on the Japanese SVCs. In total, it seems that there are 5 possible combinations of SVCs depending on their argument structure. All the 5 combinations confirm all of the properties typical for SVCs, such as no overt marker, subject sharing, tense sharing, and polarity sharing. Concerning headedness, we must bear in mind the difference between semantic and syntactic heads. There is always only one syntactic head, which does not necessarily have to be a semantic head as well. Concerning argument structures, a θ-role must be assigned to one single Subject by both verbs in the series. To sum up, so far there has not been any significant difference between SVCs in Korean and V-V structures in Japanese.

48 6 TERMINOLOGY GAPS

6.1 A Previous Approach to Serial Verb Constructions

The main objective of this section is to support the idea that so-called Serial Verb Constructions, which are found especially across Bantu languages and in Korean, can be also found in Japanese. So far, I have demonstrated their semantic and morphological properties, listed their possible combinations according to argument structure, and drawn their tree structures. The existence of SVCs in Japanese seems to be undeniable and one may wonder why this conclusion has not yet been reached. Straightforwardly, attention has been paid to these kinds of V-V constructions - however, they have been treated under a different term. In most of linguistic material written on Japanese V-V constructions, the term “compound verb” has been used.

The most prominent and influential linguist working in the field of so-called ‘compound verbs’ is Taro Kageyama. In this section, I focus on his work “Word Formation” from 200121. The attention is paid especially to one single task - to unify Kageyama’s previous approach with other works concerning SVCs. I try to eliminate the terminology gap and investigate whether Kageyama’s terminology is more accurate or if it is an older conception, and if necessary, whether it t should be revised.

6.2 A Compound Verb

A compound is a term used in linguistics for lexical items/lexemes consisting of at least two stems. A compound functions as a single unit and it is opaque to any morphological (except inflection) or syntactic operation. Compounds can be found across many languages. Furthermore, the stems, which actually form a compound,

21 Kageyama’s “Word Formation” from 2001 is actually based on his previous works from 1988 and also some previous material in Japanese. 49 can be different parts of speech and therefore many variations are possible (e.g. N+N, A+N, N+V). Depending on the language, some of these combinations are more frequent, some of them less. One possible combination, which is not very common for Germanic and Slavic languages, is a combination of two verbs V+V. Kageyama has noticed that the combination of two verbs in Japanese is very common, and he in fact asserts that there are thousands of them22. He consistently keeps using the term “compound verb” and never uses the term SVC. In this section, I will analyse in detail how he came to this concept “compound verb”.

In the “Word Formation” (2001), Kageyama seems to be less careful about the terminology than about any other part of his work. Firstly, he repeatedly uses the term “word” and keeps consulting its “hallmarks” (2001, 298). It has been known for many decades that “word” is a vague rather than precise linguistic term. Furthermore, he keeps using this vague expression when commenting on the possibilities of attaching particles in between parts of “words” and therefore breaking their boundaries (whatever their boundaries might be).

(1) *watasi-wa [inu-sae-goya]-o kat-ta. (Kageyama 2001, 299) I-Top [dog-even-house]-Acc buy-Past ‘I bought even a doghouse.’

In (1), Kageyama argues that sae cannot be inserted between the boundaries of a word23. I agree with him, but only when we speak about true compounds that are demonstrated in (1). The lexical item inu-goya meaning “doghouse” is considered to be a true compound, because, even in other languages, it has different semantics (“dog house” is not the same as a “house for a dog”), different morphology and syntax than inu no koya (‘dog’s house’). The inugoya is a compound made out of two nouns (N+N). As we can see in (1), if a lexical item is a true compound, it must be syntactically opaque and therefore the sae particle cannot be inserted. Kageyama observes: “focus particles cannot intrude into a word”(2001, 299).

22 According to Masaya Yamaguchi and his “List of Japanese Compound Verbs”, from the Japanese website “Creative Commons” (2013-07-04), there have been 3757different Japanese V-V combinations listed so far. Accessed March 24, 2015, http://csd.ninjal.ac.jp/comp/. 23 Sae, among ni, mo, and others, is one of the focus particles in Japanese, meaning ‘also’ or ‘even’. 50 Focus particles are not used between two verbs in the V-V combinations, and therefore different type of syntactic operation has to be applied in order to support or validate the idea of compound verbs.

Another type of test for defining the boundaries of a compound word is called Gapping (or backward Gapping). This test has been already tested on English compound verbs in section 2.1. Notice examples in (2) when this process of deleting one syntactic constituent is applied in order to technically avoid duplication (due to the economy of language).

(2) a. Ken-wa sukiyaki-o tabe, Naomi-wa susi-o tabe-ta. (Kageyama 1999, 299) Ken-Top sukiyaki-Acc eat Naomi-Top sushi-Acc eat-Past ‘Ken ate sukiyaki, and Naomi sushi.’

b. *Ken-wa [nomi-hazime], Naomi-wa [tabe-hazime]-ta. Ken-Top [drink-begin] Naomi-Top [eat-begin]-Past ‘Ken began to drink, and Naomi began to eat.’

In (2), the verb taberu is omitted in the first clause, because it also appears in the second clause. This syntactic operation is very common in Japanese and used on a daily basis. However, the structure in (2), when we have a combination of two V-V combinations with identical V2 - hazimeru meaning ‘begin’, is opaque to the Gapping operation. Kageyama understands this as a signal that V-V constructions are single units and therefore he prefers calling them “compounds”. However, I believe there is another reason for this resistance of Gapping test. Even though the structure signals some level of unity, it does not have to necessarily mean that we speak about compounds. Verbs can be connected on some different syntactic level. In order to gap both compound verbs and SVCs, two verbs, not only one of them, must be deleted in the second clause. This was demonstrated in section 2.1 where Gapping was tested on English compound verbs. At this point, I want to point out that the terminology “compound verb” was just the easiest solution, but which is not supported by any kind of evidence.

51 6.3 Two Types of “Compound Verbs”

Further in the Kageyama’s work, “compound verbs” are divided into two groups according to “discrepancies in semantic transparency, productivity, and ordering” (2001, 301).24 The first group of verbs, which is semantically transparent and productive is called “syntactic compounds” and seems to be formed in syntax. The second group of verbs, which is not semantically transparent and non-productive is called “lexical compounds” and seems to be formed in the lexicon.25 From now on, I propose that there is an inconsistency in terminology when something is called a compound verb and immediately said to be productive and semantically transparent.

6.3.1 “Syntactic Compound Verbs”

To begin with, there is an example of a “syntactic compound” verbs in (3).26

(3) tabe-oeru eat-finish ‘finish eating’

If we want to call these units “compound verbs”, they must have certain features typical for compounds. However, the first thing suggesting the opposite can be found in the field of semantics. Compounds, as was illustrated with the reference to the example (1) with inu-goya, should not be semantically transparent. However, the example (3) is understood as semantically transparent - the meaning of the whole construction can be understood from the meaning of its individual parts (V1

24 Notice that Lee also distinguishes between two groups of serial verbs. To compare Japanese types of “compound verbs” with Korean SVCs, see Lee (1992, 144-171). 25 There is a closer look on these two groups of compound verbs concerning whether they are or not formed in syntax or lexicon in Nishiyama (1998). 26 More examples can be found in Kageyama (2001). Some of them are as follows: “kaki-hazimeru ‘write-begin= begin to write,’ tabe-oeru ‘eat-finish = finish to eat,’ hanasi-tuzukeru ‘speak-continue = continue to speak,’ ugoki-gasu ‘move-begin = ‘begin to move,’ tabe-kakeru ‘eat-set = be about to eat,’ tasuke-au ‘help-join = help each other.’” (2001, 301) 52 and V2). It means that sequence of V1 meaning ‘to eat’ and V2 meaning ‘to finish’ simply carries the meaning ‘to finish eating’.

The second argument supporting my idea against calling these constructions by a term “compound verbs” is productivity. These ‘syntactic compounds’ seem to be very productive and for instance V2 oeru from the example (3) can be used with many different V1s. For illustration, three possible verb combinations are in (4).

(4) kaki-oeru kangae-oeru yomi-oeru write-finish think-finish read-finish ‘finish writing’ ‘finish thinking’ ‘finish reading’

When it comes to syntax, there are three processes that can interfere inside of V-V combination belonging to the group of “syntactic compounds”. All of them have been demonstrated by Kageyama himself, and the first process of “passivization” is illustrated in (5).

(5) a. korosi- kakeru (Kageyama 2001, 302) kill- be.about.to

b. koros-are-kakeru kill-Pass-be.about.to

The V-V combination korosu and kakeru in (5) belongs to the group of “syntactic compounds” and therefore passivization can be applied to the V1, as in (5). This process strongly supports the idea that this combination of verbs is not a compound. The possibility of using passivization, and probably other syntactic operations, arises from the structure of these “syntactic compounds” itself. Recall that V is actually consisting of two verbs being serialized and therefore various processes might be applied to either V1 or V2. The structure is illustrated below in (6).

(6) VP

V V

53 (7) a. utai-hazimeru (Kageyama 2001, 302) sing-begin

b. o-utai-ni nari-hazimeru Hon-sing-Dat become-begin

A second syntactic operation, illustrated in (7), weakens the idea that these V-V combinations were produced by compounding processes. We construct honorific sentences by inserting the verb naru, meaning ‘to become’, between V1 and V2. The example (7) points out that inserting of naru with a Dative marker between the sequence of V1 and V2 is possible, and consequently, once again, weakens the hypothesis of verbs being constructed by compounding.

(8) a. tabe-tuzukeru (Kageyama 2001, 302) eat-continue

b. soo si-tuzukeru so do-continue

The last syntactic operation is a proform or a substitution by using soo su-, which is also applicable in English and the process is very similar to “do so” substitution processes. This operation also supports the idea that the combination of two verbs is more likely to be originating from two different verbs being serialized than working as a single unit - as a compound verb. Soo su- proform can be applied to V1 and therefore implies that V1 itself is to some extent independent.

In this section, I have demonstrated that Kageyama’s terminology is vague and I understand it as a misconception. For the case of “syntactic compound” verbs, the term SVCs is more adequate. Although all of the above syntactic processes take place, it does not rule out any of the properties assigned to SVCs. In other words, there are no arguments against productive Japanese V-V combinations being called SVCs, but there are many arguments suggesting some inconsistency when we understand these combinations as being formed under compounding processes. Therefore, I strongly prefer revising the terminology.

54 6.3.2 “Lexical Compound Verbs”

As proposed by Kageyama, there are two types of compound verbs in Japanese - “syntactic compounds” and “lexical compounds”. In this section, I focus on the later, and try to question Kageyama’s terminology and discuss whether their structures originate from two verbs being serialized or whether they function as a single unit - a compound. To complete this task, processes tested in the previous section (6.3.1) are tested here again.

(9) omoi-tuku think-be.attached ‘to come into one’s mind’

There is a “lexical compound” verb in the example (9). 27 In the first place, I start with semantics and check if this V-V combination is semantically transparent and whether V2 is productive. The literal meaning of V1 omou is ‘to think’ and the meaning of V2 tuku is ‘to be attached’. It is truly difficult to guess the meaning of these two verbs when put together. The meaning of this combination is semantically nontransparent, or at least less transparent than with their “syntactic compounds” counterparts.

However, there are cases of ‘lexical compound” verbs when their meaning is at least partially transparent. Such cases are illustrated below in (10).

(10) a. kiki-kaesu Kageyama (2001, 301) ask-return ‘ask back’

b. nage-suteru throw-abandon ‘throw away’

27 Other examples of “lexical compound” verbs: “uti-korosu ‘shoot-kill = shoot to death,’ si-nokosu ‘do-leave = leave undone,’ oi-dasu ‘chase-take-out = send out,’ nage-suteru ‘throw-abandon = throw away,’ naki-sakebu ‘cry-shout = cry and scream’” (Kageyama 2001, 301). 55 c. naki-yamu cry-stop ‘stop crying’

Regarding productivity, there are dozens of examples of V-V combination when V2 is either kaesu, suteru, or yamu, and to some extent I consider them to be “semi productive”, however not so productive as their “syntactic compound” verbs counterparts.

As for syntax, three operations have been tested: passivization, honorification, and proform. The passivization process has been applied in the example (11).

(11) a. kaki-komu Kageyama (2001, 302) write-insert ‘fill in’

b. *kak-are-komu write-Pass-insert ‘be filled in’

According to Kageyama, V-V kaki-komu belongs to a group of “lexical compounds” and therefore it should be opaque to any kind of syntactic operations. Passivization belongs to syntactic operations and as Kageyama has presumed, this case is truly opaque to such process. Therefore, I only assume that this V-V combination works as a single unit.

(12) *o-kaki-ni nari-komu Kageyama (2001, 302) Hon-write-Dat become-insert

In the example (12), the second syntactic operation has taken place - honorification by using naru verb. This process has been already illustrated in section 4.2 and therefore I only briefly state that it really seems that so called “lexical compounds” are inert to this operation.

56 (13) a. naki-sakebu Kageyama (2001, 302) cry-shout ‘cry and scream’

b. *soo si-sakebu so do-shout

Lastly, the process of substitution or using a proform has been employed in (13). When V1 is substituted for the soo si- element, ungrammatical structures are again constructed. This again supports Kageyama’s approach for calling these V-V combinations by a term - compound.

6.4 Conclusion Concerning Kageyama’s Terminology

Whatever the terminology is, there are two types of V-V constructions in Japanese. Their structures are different, which results in different semantic and syntactic behavior. The first group, which is called “syntactic compounds” by Kageyama, is semantically transparent, very productive, and three syntactic processes can take place inside of V-V structure. Therefore I suggest that the terminology should be updated, and instead of using “compound”, the term Serial Verb Constructions, which has been established in many other languages such as Bantu or Korean, is more appropriate and should replace its earlier counterpart.

On the other hand, there does not seem to be anything wrong with using term “compound” when Kageyama speaks about “lexical compounds”. Even though some of them are semantically transparent and seem to be productive, syntactic operations remain inapplicable. Neither can passivization, honorification or proform operation be applied inside of the V-V constructions. Therefore it was suggested that SVCs and “lexical compounds” have different structures .The former arises from a two verbs being serialized, the later works as a single unit, which was formed using compounding.

Kageyama says, “…Japanese makes extensive use of compounding to produce a rich variety of complex predicates” (2001, 325). In this section, I have demonstrated

57 that besides compounding process, a serialization process is used to construct grammatical V-V sequences in Japanese.

Lastly, the existence of both compounds and SVCs, has been also demonstrated in Korean in Lee (1992). Before we move to the next section, a brief remark concerning English must be made. It seems that these two types of Japanese V-V combinations - compound verbs and SVCs, also appear in English, however with highly restricted distribution. First, the structure of Japanese compound verbs is very similar to the one in English compound verbs. Once they were constructed, they function as single units. Second, the structure of Japanese SVCs (syntactic type) also resembles the structure of English bare infinitives. The next chapter will discuss the similarity between English bare infinitives and SVCs.

58

7 A CROSS-LINGUISTIC COMPARISON IN ENGLISH, JAPANESE, KOREAN

In the very first section of this thesis, were been discussing the V-V combinations in the syntax of English. We have come to a conclusion that the structures of two successive verbs, where the V2 is appearing in the bare infinitive, satisfy all conditions typical for the phenomenon of SVCs. Furthermore, one more specification of these constructions has been concluded - there is only a limited number of English verbs that might appear in the V2 position. Similarly, in Japanese SVCs, in particular the ones that are called “syntactic compounds” by Kageyama, only limited set of verbs (often called aspectual verbs) can appear as V2.

At this point, an interesting question must come to our mind. Is it possible that these bare infinitive V-V combinations in English are same as those found in Korean and Japanese, also known as SVCs or “syntactic compounds”?

(1) Meaning Korean28 Japanese English ‘to come’ ota ? kuru come ‘to go’ kata ? iku go ‘to help’ *topta *tasukeru help ‘to run’ *tallita *hasiru run ‘to continue’ tayta tudukeru *continue ‘to keep’ nohta oku *keep ‘to give’ cwuta ataeru *give ‘to try’ pota ? miru *try ‘to complete’ perita ? simau *complete ‘to finish’ *kkuthnata owaru *finish Table 2: A Comparison of the V2 among Korean, Japanese, and English

28 Table 2 is based on the list of Korean V2 in SVCs that are listed in Lee 1992, but previously introduced in Kim (1984). 59 The Table 2 in (1) illustrates different verbs that might appear in the position of the V2. The verbs appear together with the asterisk mark because even though there is a lexical item that carries the meaning represented in the first column tagged “Meaning,” they cannot appear in the position of V2 as a part of the V-V couple. The question marks appearing in the column with the Japanese verbs mean that these verbs can appear in the V2 position, but usually when they are part of the lexical compound verbs.

(2) a. help cook rice b. run tell the neighbour

When we try to analyse the number of V2s given in the Table 2, it seems that both Korean and Japanese have more of them when compared to their English counterparts. On the other hand, there is an English verb help or run in (2) cannot be used as V2 in either Korean or Japanese. However the occurrence and the examples themselves may vary, it still seems that their structure is identical. The structure has been already demonstrated in the previous chapters and is once again illustrated below in (3).

(3) VP

V V

As previously illustrated in section 2.3 concerning English bare infinitives, there is a difference between English Bare Infinitives and SVCs in Korean or Japanese - there is no inflection on the V2 in English, but it is possible in Korean or Japanese.

7.1 Gapping

In this section the issue of the Gapping operation is discussed. This process has been already demonstrated in section 2.1 concerning English compounds and in

60 section 6.2, dealing with Japanese SVCs. It has been noted that the Gapping operation cannot take place inside either “syntactic” or “lexical” Japanese V-Vs. In order to examine the similarities or differences between SVCs and English bare infinitives, the Gapping operation will be tested in this section on English syntactic structures and similarly on Korean.

(4) John has written the words, and Paul (*has) the music. (Jackendoff 1971, 23)

It seems that there are many restrictions for the gapping operation. One of the cases is when an auxiliary verb is gapped. This case is illustrated in the example (4). It seems that both lexical verb and auxiliary verb must be deleted when the Gapping process takes place. This signals syntactic relation between the auxiliary verb and the lexical verb.

(5) Bob tried to wash himself, and Mary to read the funnies. (Jackendoff 1971, 24)

(6) Max seemed to be trying to begin to love Harriet, and Fred (((to be trying) to begin) to love) Sue. (Jackendoff 1971, 25)

It is interesting to notice how the Gapping deals with English infinitives (5). It seems that this operation is allowed even when infinitives appear. Jackendoff demonstrates that any number of infinitives can be gapped (1971, 24). This process is illustrated in the example (6).

(7) a. They go visit their mother tomorrow. b. We go visit our father on Sunday. c. *They go visit their mother tomorrow, and we go our father on Sunday. d. They go visit their mother tomorrow, and we go our father on Sunday. e. They go visit their mother tomorrow, and we go take out our father on Sunday.

Speaking about the infinitives, the question whether English bare infinitives can be gapped arises. The example (7) illustrates that Gapping cannot be applied to the

61 English bare infinitives. In order to form grammatical sentences, we can either omit both verbs from the bare infinitive series (7), or we have to use both of them (7). This signals some level of unity in these constructions and points out another similarity between SVCs and English bare infinitives. It seems that English bare infinitives, together with the V-V couple in the SVCs, are not derived from biclausal structures. Unlike English infinitives (semi clauses) that actually consist of two individual clauses, bare infinitives seem to be more consistent, which results in the restriction in the Gapping process.

(8) a. Nay-ga pang-ey kulim-ul kel-e noh-ass-ta. Korean I-Nom wall-on picture-Acc hang-L put.on-Past-Dec ‘I hung the picture on the wall.’

b. Nay-ga keul-ey sacin-ul kel-e noh-ass-ta. I-Nom mirror-on photo-Acc hang-L put.on-Past-Dec ‘I hung the photo on the mirror.’

c. *Nay-ga pang-ey kulim-ul kel-e, sacin-ul kel-e noh-ass-ta. I-Nom wall-on picture-Acc hang-L photo-Acc hang-L put.on-Past-Dec ‘I hung the picture on the wall and the photo on the mirror.’

d. Nay-ga pang-ey kulim-ul kel-e, sacin-ul kel-e noh-ass-ta. I-Nom wall-on picture-Acc hang-L photo-Acc hang-L put.on-Past-Dec ‘I hung the picture on the wall and the photo on the mirror.’

The Gapping operation has been applied on SVCs in Korean in (8). The Gapping process cannot be applied on Korean SVCs because even though the expression is consisting of two lexical verbs, neither of them can be omitted for the very same reason as with English bare infinitives. Both or neither of the verbs must be used in the first clause in order to form correct sentences.

In section 6.2 concerning the Gapping operation, it has been illustrated that Gapping cannot be applied inside Japanese V-V series (either “syntactic” or “lexical” type). Later, in this section, this operation was tested similarly on English bare infinitives and Korean SVCs. It has been demonstrated that when the Gapping process took place, the V-V couples behaved as a syntactic unit. In other words, it is

62 necessary to delete both of the verbs when gapping takes place. When only the V2 is deleted, ungrammatical structures are generated.

7.2 Stress Assignment

The Stress Assignment has previously tested on English compounds in 2.1. This section provides an analysis of stress on English bare infinitives.

(9) a. She will go ‘find her friends. b. You can come ‘drink after that. c. He will run ‘tell her his phone number.

In (9), the examples of English bare infinitives are illustrated. Unfortunately, this field has not been examined in Chomsky and Halle’s work from 1968. However, it seems that the stress always falls on the right item and therefore this structure behaves distinctly from the structure of compound verbs. Bare infinitives behave according to the Nuclear Stress Rule, which is only valid for the syntactic structures. This fact supports the idea that English bare infinitives differ from English compound verbs.

(10) ‘Compound accent rule’ in Japanese: In a compound [A B], A loses accent (if it had one originally), and B gets the compound accent (even if B does not have one originally)

A similar stress assignment can be found even in Japanese. The definition in (10) has been stated by another linguist - San Danmu (2007, 9). This definition corresponds with the English Compound Rule, where stress is located on the

63 second lexical item from the series when the item is a true compound.29 There are many examples of N-N compounds where the stress shift has taken place.

(11) a. sha’kai +se’edo → shakai-se’edo Danmu (2007, 7) society-system ‘social system’ b. kyo’oiku +se’esaku → kyooiku-se’esaku education-policy ‘educational policy’

The examples in (11) validate the rule stated by Danmu, according to which only one stress is assigned in a compound. On the left, there are examples of syntactic structures where two nouns have two stresses. On the right, there are compounds behaving as units, and therefore only one stress is assigned. The same rule should also work for compounds consisting of two verbs. Unfortunately, there has not been done any research concerning SVCs and their stress assignment.

(12) SVCs (Syntactic Type) a. yomu-dasu → ‘yomidasu read-take out ‘start to read’ b. kaki-dasu → ‘kakidasu write-take out ‘start to write’

(13) Compounds (Lexical Type) a. omoi-dasu → omoi’dasu think-take out ‘recall’ b. tutuki-dasu → tutuki’dasu bind-take out ‘factor out’

The stress assignment in Japanese SVCs and compound verbs has been tested in (12) and (13). The examples consisting of the verb dasu, which can be used in the V2 position when constructing both compound and serial verbs, has been tested. Wen the verb dasu is used as a V2 in SVCs (syntactic type), it never takes stress

29In English it is the first element that takes the stress. Due to the opposite word order of Japanese and English, the rule only seems to be inverted. 64 (12). On the other hand, when it is used in a true compound verb , it always takes the stress (13).

In this section, the analysis concerning stress assignment has been done. It has been clearly demonstrated that stress assignment is different between English bare infinitives and English compound verbs. Similarly, the same conclusion has been made on Japanese SVCs and Japanese compound verbs. The outcome of this analysis has supported the idea that English bare infinitives, together with Japanese SVCs, are similar syntactic units different from compounds.

7.3 Conclusion Concerning English Bare Infinitives

In this section two linguistic tests concerning the behaviour of English bare infinitives and SVCs have been provided.

First, the gapping operation has been tested on English bare infinitives and similarly on the Korean and the Japanese SVCs. Both English bare infinitives and SVCs behaved identically. It has been demonstrated that when the Gapping process takes place, the V-V couples behave as a syntactic unit. In other words, it is necessary to delete both of the verbs when gapping is applied. When only the V2 is deleted, ungrammatical structures are generated. These V-V series represent some level of unity on a syntactic level.

Second, the assignment of stress on English bare infinitives and compound verbs has been tested. As the outcome of this testing, it has been illustrated that these structures are different in terms of stress assignment. The idea was supported by Chomsky and Halle’s work from 1968. The tests on Japanese compound verbs and Japanese SVCs has led to the results similar to those in English. This fact again supports the idea that bare infinitives/SVCs are different from compound verbs and similar to each other.

65 Lastly, the question whether SVCs are identical to bare infinitives in English syntax is definitely difficult to answer. To fully answer it, additional data and further research concerning these verbs is necessary. Besides that, additional syntactic operations that could be tested on all of the three languages are necessary in order to fulfil this task. For this reason, this issue will be left open for further research.

66 8 SUMMARY

The objective of this thesis has been to give a detailed analysis of the Serial Verb Constructions (SVCs). This analysis has been done on three different languages - English, Japanese, and Korean. All of the objectives proposed in the Introduction section have been accomplished.

The following objectives have been accomplished:

§ Although the definition of SVCs may vary among different linguists, there are properties of SVCs, which are common to every language that is said to have these constructions. The properties are:

o there is no overt marker between the sequence of verbs o subject sharing is obligatory (object sharing is obligatory only in the case of two transitive verbs) o there is only one tense value for the sequence of verbs o negation scope takes place over the whole construction o there is only one head in the V-V couple

§ the idea that SVCs might be, to some extent, found also in English has been suggested. Due to the insufficient data, this part has not been done thoroughly and it has left the issue open for further research. However, concerning English bare infinitives (go, come, help, run), there are undeniable similarities when they are compared with SVCs. Both structures have shown similar behaviour when for example gapping and stress assignment were tested.

§ based on the contrastive analysis of Korean SVCs, it has been demonstrated that Japanese is also one of the languages, which have SVCs. Consequently, these structures have been analysed, and it has been demonstrated that they validate all the properties typical for SVCs. Later, the full list of all possible

67 combinations of verbs in V-V constructions, based on their transitivity, has been made. Regarding transitivity, it seems that there are 5 different possible combinations of V-V in Japanese: Intransitive V1 + Intransitive V2; Intransitive V1 + Transitive V2; Transitive V1 + Transitive V2; Transitive V1 + Intransitive V2; Transitive V1 + Ditransitive V2.

§ a proposal that the traditional approach to Japanese V-V constructions is out- dated has been made. It has been noted that there are two different structures of V-V series in Japanese:

o “syntactic compounds” (=SVCs) where beside inflection, other various syntactic operations can be applied. o “lexical compounds” that function as true compounds and therefore no syntactic operations can be applied on the sequence of verbs (except inflection on the head of a compound).

Their structure is therefore different and it has been suggested that there should be two different terms for these structures: “Serial Verb Constructions” and “compound verbs.“

68 9 ČESKÉ RESUMÉ

Tato práce se zabývá fenoménem sériových sloves. Sériová slovesa jsou velmi zajímavá syntaktická struktura, která se skládá z různých kombinací sloves. Přestože je tento termín běžný pro mluvčí některých jazyků, určitě jsou i takoví mluvčí, kteří s takovým termínem nikdy v kontaktu nebyli. Z tohoto důvodu je první část této práce více zaměřena na teoretické hledisko, kde jsou znázorněny základní pojmy a pojem tohoto fenoménu. Později se dostaneme k trochu konkrétnějším přístupům, které jsou doloženy odpovídajícími příklady z různých jazyků.

V této práci jsou navrženy tyto cíle:

• I když se definice sériových sloves mohou mezi lingvisty lišit, existují některé vlastnosti sériových sloves, které jsou společné pro každý jazyk, který tyhle struktury obsahuje.

• Poukázat na to, že takzvané sériová slovesa můžeme do jisté míry nalézt i v angličtině, a to v holých infinitivech.

• Na základě kontrastivní analýzy korejských sériových sloves se dokáže, že v japonština je také jeden z jazyků, které sériová slovesa mají. Dále se pokusí analyzovat tyto struktury na základě jejich tranzitivity a najít všechny možné kombinace sloves se stavbou V-V.

• Dokázat, že tradiční přístup k japonským V-V konstrukcím je zastaralý, což vyústilo v nepřesnou terminologii. Proto by terminologie týkající se složených/sériových sloves měla být aktualizována.

Tato práce se skládá z pěti tematických částí. Nejdřív jsou sériová slovesa uvedena a definována trochu obecněji. Aby bylo možné pochopit, jak tyto struktury

69 vypadají a jak fungují, příklady různých sekvencí sloves jsou znázorněny, jako je možné vidět třeba na příkladech anglicky složených sloves (Lieber 1983). Po přezkoumání sérií sloves v angličtině, se více zaměříme na sériová slovesa, jež se nachází v korejštině. Tu se zaměříme na srovnání jejich morfologických, syntaktických a sémantických vlastností (Baker 1989, Collins 1993, Lee 1992, mimo jiné). Jako výsledek této srovnávací studie, uvedeme několik univerzálních vlastnosti sériových sloves s důrazem na sdílení předmětu či podmětu (Lee 1992). Třetí část se zaobírá myšlenkou, že sériová slovesa lze nalézt také v japonštině. Pro splnění tohoto úkolu jsou na japonských slovesech ukázány typické vlastnosti sériových sloves. Čtvrtá část práce spojuje všechny již zmíněné části a otevírá nové téma předchozích přístupů k japonským V-V konstrukcím, zejména s ohledem na jejich terminologii. Zvláště se zajímáme o výrazy - „sériová slovesa“ a „sloučená slovesa.“ Poslední část je zaměřena na anglické holé infinitivy, které porovnává s již zmíněnými japonskými a korejskými sériovými slovesy.

Výsledky této práce jsou následující:

• I když se definice sériových sloves mohou na základě lingvistů lišit, existují určité vlastnosti, které jsou stejné v každém jazyku. Tyto vlastnosti jsou: o mezi sekvencí sloves se nenachází žádný afix o podmět je sdílen oběma ze série sloves (u předmětu je to povinné pouze v případě dvou tranzitivních sloves) o slovesa mají stejný čas o slovesa mají jenom jednu polaritu o jenom jedno ze sloves funguje jako „head“

• Na základě kontrastivní analýzy korejských sériových sloves bylo prokázáno, že japonština je také jeden z jazyků, které mají sériová slovesa. V důsledku toho byly tyto struktury analyzovány a bylo prokázáno, že splňují všechny vlastnosti typické pro sériová slovesa. Na základě jejich tranzitivity byl vytvořen úplný seznam všech možných kombinací sloves se stavbou V-V. Pokud jde o tranzitivitu, zdá se, že v japonštině existuje 5 různých možných kombinací V-V: netranzitivní

70 V1 + netranzitivní V2; netranzitivní V1 + tranzitivní V2; tranzitivní V1 + tranzitivní V2; tranzitivní V1 + netranzitivní V2; tranzitivní V1 + V2 ditranzitivní.

• Bylo navrhnuto, že tradiční přístup k japonským V-V konstrukcím je zastaralý. Dále bylo prokázáno, že existují dvě různé struktury V-V konstrukcí v japonštině: o „syntaktické sloučeniny“ (=SVCs), u kterých je možné aplikovat různé syntaktické operace. o „lexikální sloučeniny“, které fungují jako skutečné sloučeniny (compounds) a proto žádné syntaktické operace nemůžou být aplikovány (s výjimkou ohýbání). Jejich struktura se tedy liší, a proto by měly existovat dva různé pojmy pro tyto konstrukce: „složené slovesa“ a „sériová slovesa.“

• Posledním výsledkem této práce je poznatek, že sériová slovesa by mohla být do určité míry také nalezena v angličtině. Vzhledem k nedostatku údajů, nebyla tato část provedena dokonce. Tímto vznikl potenciál pro další výzkum. Nicméně, pokud jde o anglické holé infinitivy (go, come, help, run), jejich vlastnosti a struktura jsou nepopíratelně podobné s těmi, které byly prokázány u sériových sloves. Sériová a složená slovesa se lišili při testech s umístněním přízvuku a “gapping” testu.

71 10 REFERENCES

10.1 Works Cited

Baker, Mark. 1989. “Object Sharing and Projection in Serial Verb Constructions.” Linguistic Inquiry 20:513−533.

Bloomfield, Leonard. 1933. Language. Chicago: The University of Chicago Press.

Chomsky, Noam and Morris Halle. 1968. The Sound Pattern of English. New York: Harper & Row

Collins, Chris. 1993. “Topics in Ewe Syntax.” Ph. D. diss.. MIT. Cambridge.

Jackendoff, Ray S. 1971. “Gapping and Related Rules.” Linguistic Inquiry 2: 21-35.

Kageyama, Taro. 2001. “Word Formation.” The Handbook of Japanese Linguistics 297−325. DOI: 10.1111/b.9780631234944.2001.00012.x.

Kang, Myung-Yoon. 1988. “Topics in Korean Syntax: Phrase Structure, Variable Binding and movement.” Ph. D. diss.. Massachusetts Institute of Technology.

Kang, Seung-Man. 1997. “A Comparative Analysis of SVCs and Korean V-V Compounds” Paper presented at the 40th Anniversary of Generativism, University of Florida.

Kayne, Richard S. 1994. The Antisymmetry of Syntax. Linguistics Inquiry Monograph Twenty-Five. MIT Press

Lee, Sookhee. 1992. “The Syntax and Semantics of Serial Verb Constructions.” Ph. D. diss.. University of Washington.

72 Lieber, Rochelle. 1983. “Argument Linking and Compounds in English.” Linguistic Inquiry 14: 251−285.

Nishiyama, Kunio. 1998. “V-V Compounds as Serialization.” Journal of East Asian Linguistics 7:175−217.

Schachter, Paul. 1974. “A Non-Transformational Account of Serial Verbs. Studies in African Linguistics 5:253-270.

Song, Gihoon. 2001. “The Syntax of Serial Verb Constructions.” M.A. diss.. University of Durham.

Williams, Kay. 1963. “The Syntax of Verbs of Motion in Ijo.” Journal of African Languages 2.2:150-154.

10.2 Works Consulted

Baker, Mark C., and Stewart T. Osamuyimen. 2002. “The Serial Verb Constructions without Constructions.” Unpublished ms, Rutgerts University.

Kageyama, Taro. 1993. Bunpō to go-keisei. Tokyo: Hitsuji Shobō.

Kim, Myung-Hee. 1984. Kwuke Tonsakwu Kwunsunget Natananun Uymikwankey Yenkwu (A Study of the Semantic relations in Korean Verb Constructions), Ph. D. diss.. Ewha Women’s University, Seoul.

Lee, Juwon. 2010. Argument Composition in Korean Serial Verb Constructions.

Matsumoto, Yo. 1996. Complex Predicates in Japanese: A Syntactic and Semantic Study of the Notion Word. Tokyo: Kuroshio.

73 Rizzi, Luigi. 1978. “A Restructuring Rule in Italian Syntax,” in Keyser (ed.) Recent Transformation Studies in European Languages, The MIT Press, 113-158. Cambridge, Mass.

Takahashi, Naoko. 2005. “Grammaticalization with Saru-compounds in Japanese.” In Daijirin, edited by Matsumusa, Akira, 125-136. Tokyo: Sanseido.

Takahashi, Naoko. 2005. “Syntactic Complexity of Komu-compounds in Japanese.” In Daijirin, edited by Matsumusa, Akira, 125-136. Tokyo: Sanseido.

74

75